Pablo Hernandez Cerdan
Verified Expert in Engineering
Software Engineer and Developer
Torre-Pacheco, Spain
Toptal member since April 7, 2022
Pablo holds a Ph.D. in physics and is a top software engineer contributing to open-source libraries in image analysis and visualization. He loves the thrill of creating inspiring projects with his software skills. Pablo is proficient in C++, Python, and CMake. He can be found maintaining the release cycle of an open-source library in C++ with Python bindings or using PyTorch to create industry-level AI solutions in computer vision. He is also a contributor to a Web3 open-science journal.
Portfolio
Experience
Availability
Preferred Environment
Linux, Vim Text Editor, Tmux
The most amazing...
...project I've developed is an AI solution for a health startup for automatic segmentation of vasculature from medical images, including data curation.
Work Experience
Software Engineer
Innolitics, LLC
- Developed an app from end-to-end, later approved by the FDA, to detect enlarged hearts from CT images.
- Architected tools to handle and curate annotated data.
- Created automatic deployment scripts to AWS ECR using Docker.
Senior Deep Learning Engineer
Mundohuevo
- Developed a generative adversarial network to create images of human irises. Trained an image translation neural network to generate photo-realistic results from low-resolution images sent by clients.
- Supervised acquisition, curation, and preprocessing of the custom, private dataset. Mentored image technicians.
- Performed training on the generative model on AWS.
Deep Learning Analyst
AANappDay DMCC
- Developed core algorithms to solve imperfect information games. This included research of the game theory and machine learning community and its implementation in the C++ codebase with Python bindings.
- Performed the architectural reorganization of the whole codebase, adding CMake files to manage compilation and testing for multiplatform. Also, automatized it via CMake the injection of third-parties dependencies.
- Added CI/CD pipelines with GitHub Actions to continuously test it in Windows and Linux.
- Tracked and fixed bugs using Jira and Github platforms.
- Modernized a legacy codebase with best C++ practices. Detecting bugs and increasing the correctness of algorithms.
Senior Image Processing and 3D Modeling Engineer
UNIFi3D (Hong Kong) Limited
- Explored the research space for alternatives and identified the best solution for the client.
- Implemented algorithmic solutions based on texture manipulation.
- Implemented a deep learning (GAN) solution tailored to the client's needs.
Senior AI Engineer
Cella Medical Solutions
- Developed an AI solution for automatic segmentation of vasculature from CT images.
- Curated and organized the company datasets. Created a script to extract images with certain features for training purposes.
- Mentored co-workers on image analysis best practices, such as keeping the geometry metadata in sync with the data augmentation transformations.
Research and Development Engineer
Medizinische Hochschule Hannover
- Developed a watershed on meshes algorithm to segment alveoli in high resolution pulmonary images.
- Provided Python bindings to the high performance C++ code.
- Mentored research students on how to use it, providing extensive documentation and tests.
Senior Data Specialist
NumFOCUS, Inc.
- Received a NumFOCUS small development grant to modernize the Insight Journal.
- Provided DOI's to every article ever submitted to the Insight Journal through Crossref's API.
- Uploaded all data to the Interplanetary File System (IPFS), to meet the Web 3.0 and keep all the data reachable for reproducibility purposes.
Research and Development Engineer
Medizinische Hochschule Hannover
- Developed an SGGEN algorithm to study generations (branching) on vascular trees from pulmonary 3D images.
- Integrated into an open-source library SGEXT as an external module. Added Python bindings, testing, and CI/CD pipelines.
- Tested real images and published a research article about it.
Research and Development Engineer
CNRS
- Provided Python bindings to the digital geometry tools and algorithms Library (DGtal), a big C++ library on digital topology using PyBind11.
- Provided extensive documentation and test frameworks for other developers to add more functionality.
- Created CI/CD pipelines from scratch using Azure pipelines.
Research and Development Engineer
MacDiarmid Institute
- Developed algorithms for simulating in-silico biopolymer networks for reservoir computing (AI). The algorithms included dynamics of single polymers and force propagation on the system.
- Created Python bindings of the high performance C++ code for easy adoption.
- Created the open-source SGEXT library. Tested and documented the work, including regular releases to PyPI.
Research and Development Intern
Kitware
- Applied Wavelets algorithms to detect cracks on teeth from medical images.
- Created a custom Slicer3D application: SlicerSALT to bundle important algorithms on shape analysis research.
- Mentored by top engineers in the field of medical analysis and learned best practices on software development from industry experts.
Experience
SGEXT
The library's main purpose is to skeletonize tubular structures from input images and extract a graph representation from them. This allows characterizing the objects of interest in a simplified yet insightful way. It also provides simulation of polymer networks as spring networks and single-chain dynamics using Monte Carlo methods. The library has been used to study the statistical properties of polymeric materials and extract graphs from vasculature from medical images (airways). It will also be used to study Reservoir Computing (AI) in in-silico polymer networks.
I was the sole developer of this project. It's fully tested, documented, and with CI/CD pipelines. It provides Python bindings using PyBind11 for an easier interface in Python for researchers. The package is uploaded to PyPI regularly.
Insight Journal
https://www.insight-journal.org/The DOI provides authors extra pay for their work because they can be parsed by Web of Science, Google Scholar, etc., boosting their research statistics.
My work focused on the back-end, cleaning and curating the original database and creating a new static and open database based on GraphQL. I created automated scripts to register the current articles to Crossref through their API. The PDFs and source code of the publications were uploaded to the IPFS network. This was the first scientific journal using Web3.0 tools.
Vasomaly
https://vasomaly.com/Education
Ph.D. in Biophysics
Massey University - Palmerston North, New Zealand
Master's Degree in Biophysics
Universidad Autonoma de Madrid - Madrid, Spain
Bachelor's Degree in Physics
Universidad de Murcia - Murcia, Spain
Certifications
Introduction to Machine Learning in Production
Coursera
Deep Neural Networks with PyTorch
Coursera
Skills
Libraries/APIs
PyTorch, VTK, Python, NumPy, Scikit-learn, OpenCV, OpenGL, OpenMP
Tools
Vim Text Editor, Git, CMake, ITK, Tmux, Scikit-image, Jekyll
Languages
C++, Python, SQL, GraphQL, XML, Fortran
Platforms
Linux, Azure, Docker, AWS
Frameworks
Boost, Qt Development, Google Test
Paradigms
Wavelets
Storage
PostgreSQL, Google Cloud Development
Other
Critical Thinking, PyBind11, Artificial Intelligence, Image Analysis, Graphs, Research, Medical Imaging, Computer Vision, Deep Learning, Open Source Development, Neural Network, Algorithms, Machine Learning, Fourier Analysis, Data Science, Physics, Simulated Annealing, Physics Simulations, CI/CD Pipelines, Image Processing, Graph Theory, Graphic Design, Minimum Viable Product (MVP), Biotechnology, Technical Writing, Monte Carlo Simulations, Data Engineering, Convolutional Neural Networks (CNNs), Signal Processing, Generative Models, Object Detection, Proof of Concept (POC), Production, APIs, IPFS, GatsbyJS, Google Software, DGtal, SGEXT, SlicerSALT, Point Clouds, Machine Learning Operations (MLOps), Leadership, Generative Adversarial Networks (GANs), Supervised Machine Learning, Generative Artificial Intelligence (GenAI)
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring